关于为什么 Valgrind 显示英特尔的 MKL LAPACKE 内存泄漏的见解答案

【问题标题】：Insight as to why Valgrind shows memory leak for Intel's MKL LAPACKE关于为什么 Valgrind 显示英特尔的 MKL LAPACKE 内存泄漏的见解
【发布时间】：2016-07-11 21:16:36
【问题描述】：

这是我第一次使用英特尔的 MKL LAPACKE 和 Valgrind。不幸的是，我遇到了一些我几乎没有经验的错误。我可以就如何解决潜在的内存泄漏提出一些建议。我正在使用 Intel 的 MKL 库，所以我高度认为这个问题是我的错，但我不确定要查找什么或如何调试问题。

这是我在 MKL 安装中测试的示例文件

/*******************************************************************************
* Copyright 2010-2016 Intel Corporation All Rights Reserved.
*
* The source code,  information  and material  ("Material") contained  herein is
* owned by Intel Corporation or its  suppliers or licensors,  and  title to such
* Material remains with Intel  Corporation or its  suppliers or  licensors.  The
* Material  contains  proprietary  information  of  Intel or  its suppliers  and
* licensors.  The Material is protected by  worldwide copyright  laws and treaty
* provisions.  No part  of  the  Material   may  be  used,  copied,  reproduced,
* modified, published,  uploaded, posted, transmitted,  distributed or disclosed
* in any way without Intel's prior express written permission.  No license under
* any patent,  copyright or other  intellectual property rights  in the Material
* is granted to  or  conferred  upon  you,  either   expressly,  by implication,
* inducement,  estoppel  or  otherwise.  Any  license   under such  intellectual
* property rights must be express and approved by Intel in writing.
*
* Unless otherwise agreed by Intel in writing,  you may not remove or alter this
* notice or  any  other  notice   embedded  in  Materials  by  Intel  or Intel's
* suppliers or licensors in any way.
*******************************************************************************/
/*
   LAPACKE_chesv Example.
   ======================
   The program computes the solution to the system of linear equations
   with a Hermitian matrix A and multiple right-hand sides B,
   where A is the coefficient matrix:
   ( -2.90,  0.00) (  0.31,  4.46) (  9.66, -5.66) ( -2.28,  2.14)
   (  0.31, -4.46) ( -7.93,  0.00) (  9.55, -4.62) ( -3.51,  3.11)
   (  9.66,  5.66) (  9.55,  4.62) (  0.30,  0.00) (  9.33, -9.66)
   ( -2.28, -2.14) ( -3.51, -3.11) (  9.33,  9.66) (  2.40,  0.00)
   and B is the right-hand side matrix:
   ( -5.69, -8.21) ( -2.83,  6.46)
   ( -3.57,  1.99) ( -7.64,  1.10)
   (  8.42, -9.83) ( -2.33, -4.23)
   ( -5.00,  3.85) (  6.48, -3.81)
   Description.
   ============
   The routine solves for X the complex system of linear equations A*X = B,
   where A is an n-by-n Hermitian matrix, the columns of matrix B are
   individual right-hand sides, and the columns of X are the corresponding
   solutions.
   The diagonal pivoting method is used to factor A as A = U*D*UH or
   A = L*D*LH, where U (or L) is a product of permutation and unit upper
   (lower) triangular matrices, and D is Hermitian and block diagonal with
   1-by-1 and 2-by-2 diagonal blocks.
   The factored form of A is then used to solve the system of equations A*X = B.
   Example Program Results.
   ========================
 LAPACKE_chesv (row-major, high-level) Example Program Results
 Solution
 (  0.22, -0.95) ( -1.13,  0.18)
 ( -1.42, -1.30) (  0.70,  1.13)
 ( -0.65, -0.40) (  0.04,  0.07)
 ( -0.48,  1.35) (  1.15, -0.27)
 Details of factorization
 (  3.17,  0.00) (  7.32,  3.28) ( -0.36,  0.06) (  0.20, -0.82)
 (  0.00,  0.00) (  0.03,  0.00) ( -0.48,  0.03) (  0.25, -0.76)
 (  0.00,  0.00) (  0.00,  0.00) (  0.30,  0.00) (  9.33, -9.66)
 (  0.00,  0.00) (  0.00,  0.00) (  0.00,  0.00) (  2.40,  0.00)
 Pivot indices
     -1     -1     -3     -3
*/
#include <stdlib.h>
#include <stdio.h>
#include "mkl_lapacke.h"
/* Auxiliary routines prototypes */
extern void print_matrix( char* desc, MKL_INT m, MKL_INT n, MKL_Complex8* a, MKL_INT lda );
extern void print_int_vector( char* desc, MKL_INT n, MKL_INT* a );
/* Parameters */
#define N 4
#define NRHS 2
#define LDA N
#define LDB NRHS
/* Main program */
int main() {
    /* Locals */
    MKL_INT n = N, nrhs = NRHS, lda = LDA, ldb = LDB, info;
    /* Local arrays */
    MKL_INT ipiv[N];
    MKL_Complex8 a[LDA*N] = {
       {-2.90f,  0.00f}, { 0.31f,  4.46f}, { 9.66f, -5.66f}, {-2.28f,  2.14f},
       { 0.00f,  0.00f}, {-7.93f,  0.00f}, { 9.55f, -4.62f}, {-3.51f,  3.11f},
       { 0.00f,  0.00f}, { 0.00f,  0.00f}, { 0.30f,  0.00f}, { 9.33f, -9.66f},
       { 0.00f,  0.00f}, { 0.00f,  0.00f}, { 0.00f,  0.00f}, { 2.40f,  0.00f}
    };
    MKL_Complex8 b[LDB*N] = {
       {-5.69f, -8.21f}, {-2.83f,  6.46f},
       {-3.57f,  1.99f}, {-7.64f,  1.10f},
       { 8.42f, -9.83f}, {-2.33f, -4.23f},
       {-5.00f,  3.85f}, { 6.48f, -3.81f}
    };
    /* Executable statements */
    printf( "LAPACKE_chesv (row-major, high-level) Example Program Results\n" );
    /* Solve the equations A*X = B */
    info = LAPACKE_chesv( LAPACK_ROW_MAJOR, 'U', n, nrhs, a, lda, ipiv,
            b, ldb );
    /* Check for the exact singularity */
    if( info > 0 ) {
        printf( "The element of the diagonal factor " );
        printf( "D(%i,%i) is zero, so that D is singular;\n", info, info );
        printf( "the solution could not be computed.\n" );
        exit( 1 );
    }
    /* Print solution */
    print_matrix( "Solution", n, nrhs, b, ldb );
    /* Print details of factorization */
    print_matrix( "Details of factorization", n, n, a, lda );
    /* Print pivot indices */
    print_int_vector( "Pivot indices", n, ipiv );
    exit( 0 );
} /* End of LAPACKE_chesv Example */
/* Auxiliary routine: printing a matrix */
void print_matrix( char* desc, MKL_INT m, MKL_INT n, MKL_Complex8* a, MKL_INT lda ) {
    MKL_INT i, j;
    printf( "\n %s\n", desc );
    for( i = 0; i < m; i++ ) {
        for( j = 0; j < n; j++ )
            printf( " (%6.2f,%6.2f)", a[i*lda+j].real, a[i*lda+j].imag );
        printf( "\n" );
    }
}
/* Auxiliary routine: printing a vector of integers */
void print_int_vector( char* desc, MKL_INT n, MKL_INT* a ) {
    MKL_INT j;
    printf( "\n %s\n", desc );
    for( j = 0; j < n; j++ ) printf( " %6i", a[j] );
    printf( "\n" );
}

Valgrind 在下一行输出info = LAPACKE_chesv( LAPACK_ROW_MAJOR, 'U', n, nrhs, a, lda, ipiv, b, ldb );

Multiple markers at this line
    - 69,664 bytes in 1 blocks are possibly lost in loss record 11 of 11 [PID:
     5535]
    - 288 bytes in 1 blocks are possibly lost in loss record 9 of 11 [PID: 5535]
    - 192 bytes in 1 blocks are possibly lost in loss record 6 of 11 [PID: 5535]
    - 192 bytes in 1 blocks are possibly lost in loss record 5 of 11 [PID: 5535]
    - 256 bytes in 1 blocks are possibly lost in loss record 8 of 11 [PID: 5535]
    - 224 bytes in 1 blocks are possibly lost in loss record 7 of 11 [PID: 5535]

无论如何，我都没有更改来自英特尔的示例文件。我只是构建并运行它。以下是 Valgrind 日志文件中的详细信息：

==5535== 192 bytes in 1 blocks are possibly lost in loss record 5 of 11
==5535==    at 0x4C2AB80: malloc (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
==5535==    by 0x5898A3D: mkl_serv_allocate (in /usr/local/INTEL/compilers_and_libraries_2016.2.181/linux/mkl/lib/intel64_lin/libmkl_core.so)
==5535==    by 0x51C3192: LAPACKE_chesv (in /usr/local/INTEL/compilers_and_libraries_2016.2.181/linux/mkl/lib/intel64_lin/libmkl_intel_lp64.so)
==5535==    by 0x400A59: main (/home/troy/Eclipse_INTEL/mkl_lapacke_gcc/Debug/../chesv_row.c:112)
==5535==
==5535== 192 bytes in 1 blocks are possibly lost in loss record 6 of 11
==5535==    at 0x4C2AB80: malloc (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
==5535==    by 0x5898A3D: mkl_serv_allocate (in /usr/local/INTEL/compilers_and_libraries_2016.2.181/linux/mkl/lib/intel64_lin/libmkl_core.so)
==5535==    by 0x5074A1E: CHESV (in /usr/local/INTEL/compilers_and_libraries_2016.2.181/linux/mkl/lib/intel64_lin/libmkl_intel_lp64.so)
==5535==    by 0x51C3491: LAPACKE_chesv_work (in /usr/local/INTEL/compilers_and_libraries_2016.2.181/linux/mkl/lib/intel64_lin/libmkl_intel_lp64.so)
==5535==    by 0x51C31F2: LAPACKE_chesv (in /usr/local/INTEL/compilers_and_libraries_2016.2.181/linux/mkl/lib/intel64_lin/libmkl_intel_lp64.so)
==5535==    by 0x400A59: main (/home/troy/Eclipse_INTEL/mkl_lapacke_gcc/Debug/../chesv_row.c:112)
==5535==
==5535== 224 bytes in 1 blocks are possibly lost in loss record 7 of 11
==5535==    at 0x4C2AB80: malloc (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
==5535==    by 0x5898A3D: mkl_serv_allocate (in /usr/local/INTEL/compilers_and_libraries_2016.2.181/linux/mkl/lib/intel64_lin/libmkl_core.so)
==5535==    by 0x51C33D9: LAPACKE_chesv_work (in /usr/local/INTEL/compilers_and_libraries_2016.2.181/linux/mkl/lib/intel64_lin/libmkl_intel_lp64.so)
==5535==    by 0x51C31F2: LAPACKE_chesv (in /usr/local/INTEL/compilers_and_libraries_2016.2.181/linux/mkl/lib/intel64_lin/libmkl_intel_lp64.so)
==5535==    by 0x400A59: main (/home/troy/Eclipse_INTEL/mkl_lapacke_gcc/Debug/../chesv_row.c:112)
==5535==
==5535== 256 bytes in 1 blocks are possibly lost in loss record 8 of 11
==5535==    at 0x4C2AB80: malloc (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
==5535==    by 0x5899F49: mm_account_ptr_by_tid..0 (in /usr/local/INTEL/compilers_and_libraries_2016.2.181/linux/mkl/lib/intel64_lin/libmkl_core.so)
==5535==    by 0x5898399: mkl_serv_allocate (in /usr/local/INTEL/compilers_and_libraries_2016.2.181/linux/mkl/lib/intel64_lin/libmkl_core.so)
==5535==    by 0x51C3192: LAPACKE_chesv (in /usr/local/INTEL/compilers_and_libraries_2016.2.181/linux/mkl/lib/intel64_lin/libmkl_intel_lp64.so)
==5535==    by 0x400A59: main (/home/troy/Eclipse_INTEL/mkl_lapacke_gcc/Debug/../chesv_row.c:112)
==5535==
==5535== 288 bytes in 1 blocks are possibly lost in loss record 9 of 11
==5535==    at 0x4C2AB80: malloc (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
==5535==    by 0x5898A3D: mkl_serv_allocate (in /usr/local/INTEL/compilers_and_libraries_2016.2.181/linux/mkl/lib/intel64_lin/libmkl_core.so)
==5535==    by 0x51C338A: LAPACKE_chesv_work (in /usr/local/INTEL/compilers_and_libraries_2016.2.181/linux/mkl/lib/intel64_lin/libmkl_intel_lp64.so)
==5535==    by 0x51C31F2: LAPACKE_chesv (in /usr/local/INTEL/compilers_and_libraries_2016.2.181/linux/mkl/lib/intel64_lin/libmkl_intel_lp64.so)
==5535==    by 0x400A59: main (/home/troy/Eclipse_INTEL/mkl_lapacke_gcc/Debug/../chesv_row.c:112)
==5535==
==5535== 69,664 bytes in 1 blocks are possibly lost in loss record 11 of 11
==5535==    at 0x4C2AB80: malloc (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
==5535==    by 0x589A184: mm_account_ptr_by_tid..0 (in /usr/local/INTEL/compilers_and_libraries_2016.2.181/linux/mkl/lib/intel64_lin/libmkl_core.so)
==5535==    by 0x5898399: mkl_serv_allocate (in /usr/local/INTEL/compilers_and_libraries_2016.2.181/linux/mkl/lib/intel64_lin/libmkl_core.so)
==5535==    by 0x51C3192: LAPACKE_chesv (in /usr/local/INTEL/compilers_and_libraries_2016.2.181/linux/mkl/lib/intel64_lin/libmkl_intel_lp64.so)
==5535==    by 0x400A59: main (/home/troy/Eclipse_INTEL/mkl_lapacke_gcc/Debug/../chesv_row.c:112)
==5535==

我应该补充一点，我正在使用 eclipse 来运行我的软件。有没有一种特殊的方法可以用 eclipse 设置 MKL 库？我相信这是我潜在的错误所在。该程序给了我一个答案，一切“似乎”都是正确的，但是在第一次运行 Valgrind 之后，我得到了上面列出的错误。有人可以提供有关如何在 Eclipse 中设置 MKL 库（路径和库）的说明吗？谢谢。

更新：我已经尝试了这些 tools 和 sites，但我仍然从 Valgrind 获得了内存反馈。在这一点上，除非有人提供有用的东西，否则我不得不假设这只是一个“潜在的”泄漏，而不是一个“确定的”泄漏。

【问题讨论】：

我相信您已经阅读了他们的FAQ，但“有问题”的输出可能是由未加载的共享对象引起的。
我现在真的是头灯里的鹿。
@Shark 你能详细说明一下吗。我搜索了该链接并阅读了该部分，但它对我来说是一门外语。
这其实是个大问题。如果您愿意，您需要一个“程序员”来查看代码并确保共享对象（库/库）被及时发布。但我建议也询问/搜索邮件列表，那里的人很有帮助。请记住，可能丢失并不意味着内存泄漏，但它可能是，只有肯定丢失意味着内存泄漏。
@Shark 你指的是 Valgrind 的邮件列表吗？您是在暗示我可能没有正确设置 MKL 库吗？任何有助于缩小我的搜索范围的事情都会有所帮助。感谢您的意见。

标签： eclipse memory memory-leaks valgrind intel-mkl

【解决方案1】：

在与英特尔的某个人取得联系后，我被建议阅读下面引用中指出的用户指南部分，并检查了此link 以进一步使用“mkl_free_buffers”。最后，添加“mkl_free_buffers();”到程序结束时解决了这个问题。 Valgrind “没有发现问题需要报告”。

避免英特尔 MKL 中的内存泄漏

When running, Intel MKL allocates and deallocates internal buffers to facilitate better performance. However,
in some cases this behavior may result in memory leaks.
To avoid memory leaks, you can do either of the following:
• Set the
  MKL_DISABLE_FAST_MM
  environment variable to 1 or call the
  mkl_disable_fast_mm()
  function.
Be aware that this change may negatively impact performance of some Intel MKL functions, especially for small problem sizes.
• Call the
  mkl_free_buffers()
  function or the
  mkl_thread_free_buffers()
  function in the current
  thread.
For the descriptions of the memory functions, see the Intel MKL Reference Manual, available in the Intel
Software Documentation Library.

【讨论】：